ASN's Mission

ASN leads the fight to prevent, treat, and cure kidney diseases throughout the world by educating health professionals and scientists, advancing research and innovation, communicating new knowledge, and advocating for the highest quality care for patients.

learn more

Contact ASN

1401 H St, NW, Ste 900, Washington, DC 20005

email@asn-online.org

202-640-4660

The Latest on Twitter

Kidney Week

Abstract: PO1030

Artificial Intelligence to Evaluate Vascular Access Aneurysms in Hemodialysis Patients

Session Information

Category: Dialysis

  • 703 Dialysis: Vascular Access

Authors

  • Zhang, Hanjie, Renal Research Institute, New York, New York, United States
  • Preddie, Dean C., Azura Vascular Care, Malvern, Pennsylvania, United States
  • Krackov, Warren S., Azura Vascular Care, Malvern, Pennsylvania, United States
  • Sor, Murat, Azura Vascular Care, Malvern, Pennsylvania, United States
  • Waguespack, Peter, Fresenius Medical Care North America, Waltham, Massachusetts, United States
  • Kuang, Zuwen, Fresenius Medical Care North America, Waltham, Massachusetts, United States
  • Ye, Xiaoling, Renal Research Institute, New York, New York, United States
  • Kotanko, Peter, Renal Research Institute, New York, New York, United States
Background

Vascular access aneurysms are a frequent finding in hemodialysis patients with arterio-venous (AV) fistulas and grafts. Of great concern is aneurysm rupture that may result in fatal hemorrhage. To that end we used artificial intelligence (AI) to automatically evaluate vascular access aneurysms.

Methods

We collected images of a diverse range of AV vascular accesses using mobile devices. Vascular access experts adjudicated the images and diagnosed the severity of AV fistula and graft aneurysms. We then randomized the images for training (70%) and validation (30%). We trained a convolutional neural network (CNN) utilizing Amazon SageMaker platform. CNN performance was measured by the area under the receiver operating characteristics (ROC) curve in the validation images.

Results

We collected 1,341 AV access images in patients dialyzed in 20 Renal Research Institute clinics in six U.S. states. The adjudication of images identified 1,093 not advanced and 248 advanced aneurysms; examples are shown in Figure 1. With the validation images, we achieved an area under the ROC curve of 0.96. Considering different probability threshold for advanced aneurysm, if threshold is 0.37, we achieved sensitivity of 80%, specificity of 95%, false positive rate of 5%, precision of 79%, if threshold is 0.7, sensitivity of 66%, specificity of 99%, false positive rate of 1%, precision of 92%.

Conclusion

Our solution of applying advanced AI technologies achieved very high sensitivity, specificity, precision, and a low false positive rate. The CNN could assist the clinical staff with actionable information and improve clinical outcomes.

Severity of AV access aneurysms. Panel A shows the images from 6 patients with not advanced AV aneurysms. Panel B shows images from 6 patients with advanced AV aneurysm.

Funding

  • Commercial Support